Slide Presentation from the AHRQ 2007 Annual Conference

Syndromic Surveillance Basics

Notes:

The main theoretical underpinning of detection is signal detection theory, which has relatively familiar concepts.

The detection problem is conceptualized as concise statement of a detection problem (or of a detection system) might include descriptions of what is to be detected, the types of input signal, and the detector itself.

Event such as an epidemic that you are trying to detect

Data that contains signal and noise.

Detector that analyzes the data, and when some THRESHOLD is exceeded produces as output an alarm

The alarms may be true or false alarms and we therefore measure the performance of detection systems using the metrics sensitivity and specificity.

By manipulating the alarm threshold, we can reduce the increase the number of true alarms by lowering the threshold but almost always at the cost of increasing the numbers of false alarms and vice versa. Usually we are not sure what mix is optimal, so we summarize it with a receiver-operating characteristic curve.

Timeliness can also be measured by comparing the time of the signal to the time of the vent itself. Timeliness can also be improved by adjusting the threshold of detection so that all three measures can be improved at the expense of the others.